Le Carrer, Noemie 
ORCID: 0000-0002-6057-2057
(2021)
A possibilistic framework for interpreting ensemble predictions in weather
forecasting and aggregate imperfect sources of information.
PhD thesis, University of Liverpool.
|
Text
201195354_Apr2021.pdf - Unspecified Download (7MB) | Preview |
Abstract
Until now, works in the field of tide routing (i.e., optimization of cargo loading and ship scheduling decisions in tidal ports and shallow seas) have omitted the uncertainty of sea level predictions. However, the widely used harmonic tide forecasts are not perfectly reliable. Consequences for the maritime industry are significant: current solutions to tide routing may be made robust through the introduction of arbitrary slack, but they are not optimal. Given the financial implications at stake for every additional centimeter of draft and the catastrophic effects of a grounding, an investigation of tide routing from the perspective of risk analysis is necessary, which we first develop in this PhD thesis. Predicting future sea level errors w.r.t. tide predictions can be achieved by statistical modelling of these errors, based on historical archives, or by physics-based numerical predictions of these deviations. In the latter option, ensemble forecasting has gained popularity in the field of numerical weather prediction as a way of quantifying the uncertainty on forecasts. Tide-surge ensemble forecasts are thus routinely produced, combining hydrodynamic models with weather ensembles. This type of forecasts is commonly interpreted in a probabilistic way. However, the latter is regularly criticized for not being reliable, especially for predicting extreme events because of the chaotic nature of the dynamics of the atmospheric-ocean system, model error, and the fact that ensemble of forecasts are not, in reality, produced in a probabilistic manner. In this PhD thesis, we consequently develop an alternative possibilistic framework to interpret and use operationally such ensembles of predictions. In particular, we show by numerical experiments on the Lorenz 96 system that probability theory is not always (e.g. at large lead times and extreme events) the best way to extract the valuable information contained in ensemble predictions. Besides, such a possibilistic perspective eases the combination of different imperfect sources of information about the future state of the system at hand (e.g. dynamical information based on past time series and the analog method), in addition to making more sense without the need of post-processing. Finally, combining both the scheduling problem and the ensemble interpretation solution, we design a shipping decision model to compute optimal cargo loading and scheduling decisions, given the time series of the fuzzy sea levels in these ports that we derive from a possibilistic interpretation of surge ensemble forecasts. The under keel clearance becomes a possibilistic constraint and the resulting shipping optimization problem is solved by means of an optimisation routine adapted to possibilistic variables. Results obtained on a realistic case study with 7-day-ahead tide surge ensemble predictions are discussed and compared with those given by a probabilistic approach, or by standard practices on ships. After our numerical case studies on the Lorenz 96 system, they illustrate the potential and limitations of a possibilistic interpretation of the weather ensemble forecasts over its probabilistic counterpart in a realistic setting.
| Item Type: | Thesis (PhD) |
|---|---|
| Divisions: | Faculty of Science and Engineering > School of Engineering |
| Depositing User: | Symplectic Admin |
| Date Deposited: | 25 Jun 2021 10:43 |
| Last Modified: | 18 Jan 2023 22:34 |
| DOI: | 10.17638/03126825 |
| Supervisors: |
|
| URI: | https://livrepository.liverpool.ac.uk/id/eprint/3126825 |
Dimensions
Dimensions
